Daniel Khashabi

World-In-World: World Models in a Closed-Loop World

IA2: Alignment with ICL Activations improves Supervised Fine-Tuning

BiomedSQL: Text-to-SQL for Scientific Reasoning on Biomedical Knowledge Bases

Jointly Reinforcing Diversity and Quality in Language Model Generations

The Alignment Waltz: Jointly Training Agents to Collaborate for Safety

Hidden in the Haystack: Smaller Needles are More Difficult for LLMs to Find

The Flaw of Averages: Quantifying Uniformity of Performance on Benchmarks

Linguistic Nepotism: Trading-off Quality for Language Preference in Multilingual RAG

Safe and Efficient In-Context Learning via Risk Control

Gold Panning: Turning Positional Bias into Signal for Multi-Document LLM Reasoning

Can A Society of Generative Agents Simulate Human Behavior and Inform Public Health Policy? A Case Study on Vaccine Hesitancy

Controllable Safety Alignment: Inference-Time Adaptation to Diverse Safety Requirements

GenEx: Generating an Explorable World

Hell or High Water: Evaluating Agentic Recovery from External Failures

SIMPLEMIX: Frustratingly Simple Mixing of Off- and On-policy Data in Language Model Preference Learning

FEEDBACK FRICTION: LLMs Struggle to Fully Incorporate External Feedback

CycleAug: Cycle-Consistent Visual Augmentation for Large Multimodal Models